Deepfake Training Only Improves Detection 10%

Nautilus might be trying to scare people with the FUD in an article called “Deepfake… Should Scare Us

The most recent study, by psychologist Sophie Nightingale of Lancaster University and computer scientist Hany Farid of the University of California, Berkeley, focused on deepfake images.1 In one online experiment, they asked 315 people to classify images of faces, balanced for gender and race, as real or fake. Shockingly, overall average accuracy was near chance at 48 percent, although individuals’ accuracy varied considerably, from a low of around 25 percent to a high of around 80 percent.

In a follow-up experiment, 219 different people did the same task but first learned about features to look for that can suggest an image is a deepfake, including earrings that don’t quite match, differently shaped ears, or absurdly thin glasses. People also received corrective feedback after submitting each answer. “The training did help, but it didn’t do a lot,” Nightingale told me. “They still weren’t performing much better than chance.” Average accuracy increased to only 59 percent.

My guess is the training to detect fakes wasn’t very good.

That’s an equally valid conclusion, which I don’t see mentioned here. What if training could be introduced that helped more?

To put it another way, does a test and education program about fakes include thinking about people who are blind and don’t trust any visuals? There’s an important clue in the article about why testing vision alone may be a poorly-contrived exercise:

Part of the reason it remains so difficult to make a believably realistic recreation of young Luke is because he has to emote and speak. Even in The Book of Boba Fett, producers clearly recognized the limits of their illusion, frequently cutting away from Luke’s face whenever he had extended dialogue.

So what are they training people on exactly, and why use such a narrow band, when a simple second factor would increase detection rates significantly?

Here’s another clue in the article, which reveals how “risk” analysis can end up exactly backwards:

We’re such expert face detectors that we can’t help but see faces everywhere: in rock formations on Mars, in the headlights and grilles of cars, and in misshapen vegetables.

Nobody seeing faces in misshapen vegetables gets an award for being an expert face detector. That’s a massive contradiction to the entire premise anyone should be scared by the fact that fake faces can look like real ones.

This all goes back to the main point I often try to make, which is society tends to very much like and enjoy fakes until it doesn’t. That has to be kept in mind.

Does going into a theater make it more acceptable to watch people fake other people (acting), as a form of containment, than if they do it on the street or in our homes? Every Halloween I welcome many (admittedly marginal quality) deep fakes into my environment and nobody seems worse for it.

When a person walking up to you says “I’m your father” there are a million data points in your mind evaluating that statement. When someone says “I’m celebrity X depicting fake character Y” there are significantly fewer points to evaluate. And if a researcher asks “Is this picture a real person” there are even fewer points.

Scary? At the end of the day social engineering is a problem yet hardly a new topic, so I often wonder why deep fakes are so exciting to people now instead of many years ago or even decades.

More than 20 years back I had to slide into environments, engineering my way to walk out with someone’s internal hard drive in my hands (setting an exact replica inside their computer instead) without anyone at a facility noticing.

Layers of presenting fake information are a professional exercise across many industries, probably not unlike the kind of medical operations we have come to accept as normal and beneficial (e.g. organ transplants to save a life, plastic surgery to repair burns).

Instead of being scared by the premise of challenging areas of weak trust scaffolding (e.g. looking at someone’s face to determine something), people need to think more broadly about what is really at stake in a society that is scared by fakes.

Here’s a more important and even simpler test:

A black woman sends messages using an undetectable appearance of white male celebrity.

Who does this example scare, and why?

1864 Concentration Camp of Andersonville, Georgia

The “Deadliest Ground of the Civil War” officially was recorded as a concentration camp run by Confederate soldiers where they systematically tortured and killed American Prisoners of War (POW).

Three days after the camp was opened with the first 400 POW, already one died. Six months later it was 3X more populated than its alleged initial design, cruelly concentrating over 45,000 people into death-camp conditions without proper shelter, food or sanitation.

Soon after, over 10,000 Americans were dead.

Andersonville was very intentionally an act of war crimes. It was ordered and used by the Confederate President to maximize suffering, consistent with the overt and formally stated white police state articles (Confederacy) of preserving and expanding slavery through violent means.

Source: Andersonville National Historic Site, NPS.gov

Some say winners write history, yet in America it often is the opposite. The obvious losers of the Civil War have polluted history in order to obscure their crimes.

Despite best efforts of Confederate-apologist historians to destroy or hide the truth — take unfair advantage of peace-time “healing” to denigrate American heroes and peddle propaganda about America’s enemies — the fairly well-documented Andersonville crimes should still stand out.

Purposeful concentration camps for the torture and murder of captured American soldiers were without question direct response to the emancipation of Black people; initiated under Confederate President Davis who terminated all prisoner exchanges.

…in September of 1862, President Lincoln called for the enlistment of black soldiers into the Union Armies as part of the preliminary draft of the Emancipation Proclamation. In December 1862, President Davis responded by issuing a proclamation that neither captured black soldiers nor their white officers would be subject to exchange.

More to the point, a Civil War started to preserve and expand slavery meant a prisoner exchange was unable to overcome fundamental belief in the South that they stood for murdering any Black person at any time (or anyone defending a Black person) without consequences.

Prisoner terms of freedom violated the very principle of the white prison states, which formed into a Confederacy to forever deny freedom to millions of its existing Black population held as prisoners in labor camps (plantations).

The South overtly refused to exchange or free any Black person captured because they did not consider Black people human. Many ruthless and immoral men in leadership positions of the South, such as the barbaric traitor General Lee, even allowed the outright execution of POW and civilians.

General Chalmers (Mississippi cavalry who later became known for using violent voter suppression to win a seat in Federal government) reportedly bragged about this event in words similar to General Lee that a butchering at Fort Pillow was intentional and to teach “the mongrel garrison” a lesson. Harper’s Weekly described the situation in their 1864 news report as murdering women, children and then mutilating the dead.

The term mongrel was meant to suggest that any Blacks captured as soldiers meant the Confederate General believed he would be justified in murdering all POW in their company, regardless of race, as a terror tactic.

Terror tactics indeed left an impression. The Andersonville concentration camp became news soon after the Confederacy lost the Civil War (which might sound familiar to anyone reading WWII narratives about Nazi death camps) and Americans promised to not forget.

Source: Hargrett Rare Book and Manuscript Library, University of Georgia Libraries.

It seems there was sufficient widespread outrage, quick documentation and public reaction to cement the facts about its atrocities in an attempt to hold someone responsible.

Following the Confederate surrender in April 1865, Clara Barton, later founder of the American Red Cross, and Dorence Atwater, a former prisoner assigned as a parolee to keep burial records for prison officials, visited the cemetery at Andersonville to identify and mark the graves of the Union dead. During the war Atwater had labeled the soldiers by name and number after their deaths. Through Barton and Atwater’s efforts, the cemetery was dedicated as Andersonville National Cemetery in August 1865.

Among the various men tried and executed for heinous crimes committed during Civil War, Captain Hartmann Heinrich Wirz became infamous as head of the Andersonville death camp.

Another man named Robert Kennedy, for example, was tried and executed in March 1865 for the crime of placing terror bombs around New York City landmarks and public spaces.

More to the point, Champ Ferguson was tried and executed in October 1865 for cruelly murdering prisoners of war.

…convicted in the fall of 1865 for the execution of at least 53 captured Union soldiers, although Ferguson claimed the total was higher.

Ferguson must roll in his grave when people ignore his story despite best attempts to exaggerate the number of POW he had murdered.

Wirz stood on the gallows as just one of the many who could have been convicted.

He hanged to death in the courtyard of the Old Capitol prison of Washington D.C. on November 10, 1865, after nearly 150 witnesses testified he had personally ordered death for POW and he was found guilty in 11 of 13 counts of documented acts of personal cruelty.

For example, one Confederate soldier testified that Wirz ordered a prisoner into the stocks during a rainstorm. The soldier, observing the prisoner was drowning, placed an umbrella over the prisoner and approach Wirz, who replied, “Let the damned Yankee drown.” […] Whether or not Wirz violated the existing laws of war is not subject to debate.

However, others responsible for war crimes including mass murder of American POW soon were to have a different fate, as the NPS explains. The Federal government abruptly halted prosecutions under the premise of “healing” the nation.

There were further military tribunals against Confederates planned in the spring of 1866. For example, a board of inquiry found that there was sufficient evidence to charge General George Pickett, of Gettysburg fame, for signing off on the execution of twenty two North Carolinians serving in the Union Army who were captured at New Bern, NC in February 1864. However, thanks to the intercession of his old West Point classmate Ulysses Grant and President Johnson’s April 1866 proclamation that the rebellion was over, Pickett was never arrested and charged by a military tribunal. Johnson’s 1866 proclamation specifically banned military tribunals in peacetime, and effectively put a stop to any further arrests and charges like those brought against Henry Wirz.

One of the peculiar details about Andersonville is how quickly it was filled and then also abruptly ended, making it unquestionably a killing field.

President Davis’ 1862 order to stop all prisoner exchange, as described above, had manifested by 1864 in building this one massive concentration area, essentially making a cramped walled field of Georgia into the fifth-largest city of the South.

Why weren’t these prisoners dispersed more widely instead? Why were prisoners so easily added to the camp, or even removed from it, yet necessary supplies were not added so easily along with them? Why were supplies still being transported so easily north into Atlanta by way of Macon Railroad yet none reached Andersonville further south on that same line?

Source: Andersonville National Historic Site, NPS.gov

An urgent order to disperse the concentration camp then suddenly came when liberating armies drew close and were poised to set its victims free.

General Sherman captured Atlanta on September 2, 1864. Word quickly reached Andersonville and mass evacuations began immediately. In just the one week of September 7-13 nearly 17,000 prisoners were transferred to other prisons in Georgia and the Carolinas. In mid-September, Sherman and Confederate General Hood negotiated a “special exchange” for those captured in the Atlanta campaign and around 2,000 prisoners were sent to Atlanta for exchange. By the end of the month less than 9,000 prisoners remained at Andersonville. When Sherman began his March to the Sea on November 15, 1864, there were less than 200 prisoners in the stockade and less than 2,000 in the hospital. That very day an additional 500 were transferred to Savannah lowering the prison’s population even further. The death count on November 15, 1864 stood at around 12,100.

In other words, the prisoners saw supply chain logistics that could have delivered supplies but they were denied. The prisoners similarly saw supply chain logistics for rapid dispersal and relocation but they were denied. The pieces of the puzzle were in place to avoid a death camp, yet the South chose death camp.

I even have seen some military historians try to claim that Sherman didn’t march on the camp because to liberate it would free more Rebels to fight. This seems exactly backwards, since the Union threatening liberation of the camp meant Rebels focused on it instead of battles elsewhere. The threat of camp prisoners being free meant the Confederate South sprang into action denying them that chance. Far more resources of the Confederacy were consumed in desperate acts to prevent POW release. When the camp had been ignored by the Union, however, the Confederates similarly had ignored it and tried to run it without even the bare minimum of staff and supplies.

Consumers Reject 2022 Apple MacBook Pro for 1981 Commodore Design

A comment today on Twitter seems completely divorced from basic computer history, and it’s gathering a lot of misplaced attention as something new.

Source: Twitter

Similar attempts at false “new” ideas can also be found on other platforms and channels.

Source: Reddit

I’ve discussed here before how Apple always has been following others around since the beginning; a constant late-mover in the personal computer industry that profits from taking the ideas of others without giving credit.

With all due respect I don’t know what new things they’ve done. […] Everything was demonstrable in 1973.

Ouch.

It’s usually trivial to prove how Apple (and its loyal users) copy someone else’s ideas and cannibalize them, while pretending to be innovative.

Consider, for example, that a highly popular very affordable Commodore VIC-20 design in 1981 for the personal computer looks very much like the allegedly new concept in 2022 — a keyboard with the computer inside and plugged into any monitor, using peripherals for expansion.

It was unquestionably a best-selling personal computer with a low price of $299 that quickly made it the first computer to reach 1 million units sold.

Apple lagged behind Commodore in sales.

Boom.

It ended up selling another 4 million before 1985. Neither obscure nor forgotten, Commodore was a true innovator and sales leader.

Moreover, Commodore saw such incredible demand for this exact design that sales ballooned into 20 million Commodore computers delivered, straining ability to keep up with demand. Subsequent models included the famous Amiga (easily one of the best personal computer products in history).

The Amiga offered both this keyboard design, and also a standalone thin “box” or even the amazing headless “toaster” model… Commodore had such ground-breaking engineering that by 1988 an Amiga was running like a personal computer from 1998, ten years ahead of its time.

Now in 1990, on the Mac side of the fence, Apple was still charging over $6,000 for a black and white Macintosh (like the SE/30). Before 1987, they couldn’t run more than one program at a time. When they finally did do multitasking, it was with a crash-prone method called co-operative multitasking. Contrary to popular myth, the first true pre-emptive 32-bit multitasking colour Mac didn’t arrive until the release of OS-X in 1999.

The PC side of the fence was far worse. For those who have never experienced the “joy” of a PC running MS-DOS refusing to boot because the AUTOEXEC.BAT or CONFIG.SYS file isn’t configured correctly, just imagine the computer equivalent of root canal surgery. It didn’t get a colour pre-emptive multitasking operating system until Windows 95.

In contrast, first released in 1985, the Amiga was a useful colour video editing tool. By 1990, you could hook up to four video cameras up to one and switch between them in real time.

By 1990 we were able to take an Amiga, a keyboard with a computer inside, plug it into any television to watch videos in color that we downloaded on the Internet while we edited our documents. The high-end models like the Amiga 3000 look like a computer 20 years ahead of its time.

Taking the integrated monitor off an Apple computer to get “likes” today on a social platform is… sad.

The market seems to constantly reward laggards and follow-on products, and too easily fails to recognize true innovators, let alone keep in mind some basic knowledge of computer history.

Wikipedia Vandalism on “Black Tom” Entry

On the 20th of February a mobile user from 73.245.188.148 (Miami, FL) made three sets of edits to a Wikipedia page about the “Black Tom” explosion in 1916.

The teenage-themed hateful vandalism edits are intentionally obvious.

Click to enlarge

Now I want to draw your attention to the speed of the revision listed at the top.

  • Revision as of 00:13, 20 February 2022 (vandalism)
  • Latest revision as of 00:19, 20 February 2022

That’s an impressive restore time of about 5 minutes. Yet it wasn’t enough to prevent search engines from absorbing the vandalism and keeping it alive much longer.

Even today — March 4th — a lack of integrity in Wikipedia still flows to users, with breached data spreading via search engines like DuckDuckGo and Qwant.

This brings up the question of who is really targeted by the vandalism, given how and where such an attack manages to impact information integrity.

Bing and Google appear unaffected.